I read that Pedro Cotter was using a PCIe Matro G550 card in his Amithlon system so I decided to test one out. I had removed all the additional framebuffer drivers from my kernel so I did a quick compile of a new one re-adding the Matrox framebuffer and ATI Radeon framebuffer (that one will be tested next). I removed the nVidia FX 5200 64MB PCI card and installed the Matrox G550 32MB PCIe card. Powered on (I had already swapped out the kernel on my DOS boot partition and edited the “small” file to reference it), and it booted just fine to my Workbench. Went in to Picasso96Mode prefs and selected the Matrox-1 card as the board in Picasso96Mode. Rebooted and all continued to work though I lost a few hi-res hi-color modes, 1650×1080 @ 32 bit and 1280×1024 @ 32 bit, leaving on the 8 bit and 16 bit modes of those resolutions. I use my own defined 1280×720 @ 32bit so no worries. The reason is the lower amount of available memory the Matrox framebuffer allows, only 4 MB versus the 16 MB you get out of an nVidia card. I had to recreate a 320×240 mode as the one I created for my nVidia card threw an “input signal out of sync” on my monitor(s). Did some regular use testing and didn’t feel much difference, and did some benchmark testing which showed a significant reduction in most test. A few scored higher, but on average most were slower, some significantly slower. I attribute this to those functions being tested were not hardware assisted. I believe nVidia has the highest amount of hardware accelerated functions enabled which gives it a performance advantage.
In real use I’m not sure it’s something you’ll notice that much if at all. Windows still open and move without feeling different, games still are fast enough. Simple Quake timedemo demo1 showed less framerate than the nVidia card, but it was still smooth and above 30 fps at 1024 x 768 @ 16bit.
This opens up a PCI slot for the Catweasel (yay) and gives more options for new(er) motherboard choices with supported chipsets, as a lot of the latest ones typically only had two PCI slots (Video and Networking or Video and Audio) and a few PCIe slots. Now you can have PCI Audio and Networking and PCIe Hardware Accelerated Video.
I plan to run benchmarks in SysSpeed for each video card and save the results as a module and make them available for download as well as displaying the overall results.
Picking up a Catweasel Mk4 PCI card, hoping this will work well with Amithlon allowing me to read/write Amiga formatted floppies. The device allows reading/writing Amiga DD/HD, PC DD/HD and Mac formatted disk with a standard PC floppy drive.
The fun part…I’m out of PCI slots. So I need to see if my PCI-e to PCI adapter for my motherboard works or perhaps try a PCI-e network card in place of the PCI one I’m using.
This will be interesting….
The Catweasel is installed! Multidisk.device copied to devs: and I created two mountlist, one for Amiga DD disk and one for Amiga HD disk. Double clicked the DD one and then inserted an Amiga DD disk and its icon popped up on my Workbench. Double clicked it and it opened the contents of the disk. I removed the floppy and the disk icon went away……only to be replaced by that familiar Amiga disk drive clicking.
Any noclick tools work with multidisk.device? – Yes, the Multidisk prefs program from the Catweasel 2 software. No click enabled….. no clicking. Bonus, the Catweasel 2 software also included a patch for CrossDOSFileSystem that makes it work with multidisk.device and supports disk changes (no diskchange ph0: or ph1:).
If you’ve followed my guide for compiling a kernel and have spent any amount of time looking at the .config file…..I’m curious if anyone has experimented with what’s really needed or not? I launched VirtualBox the other day, booted my Centos setup for compiling the Amithlon kernel and dumped one out and tested on my actual system. I made very few changes, mostly removing all the modules [m] options since I don’t use any modules. I’m more curious if some of the modules are expected to be loaded (depending on hardware) that I could be missing that would be better off compiled in to the kernel ([*] instead of [m]). Honestly the only things I remember ever using modules for were network drivers, audio drivers and/or network drivers. A lot of character/keyboard/serial selections are designated at modules in Milan’s .config file.
After booting and testing the first one I went back to Centos and stripped out all the IDE drivers, just leaving generic and Intel (my motherboard uses an Intel SouthBridge and NorthBridge) IDE, enabling SATA support and adding the drivers for a possible SCSI card (I have) if I feel like giving it a try. My kernel wasn’t much smaller than the one I compiled and have been using. I’d like to strip out more unneeded tags or drivers.
I have to admit, I was trying to make a more generic kernel while being specific at the same time…one I can use in VirtualBox as well as on my hardware system. Adding NIC and Sound drivers for options VB presents and such. I could strip those out to be leaner I suppose.
Anyone else compiling their own Amithlon kernel and discovering what can be left out and what is absolutely required?
I’ve been using a slim DVD drive that fits in a 5.25″ bracket that holds the slim drive and a 3.5″ device in my A4000T and really liked the look. I thought about adding one to my Amithlon but didn’t know what I’d put in the 3.5″ area (it can remain empty or contain a device you don’t need access if you leave the 3.5″ cover attached (it’s a break off piece)). I use a CF Card reader in my A4000T in that spot. I also have a 3.5″ USB multi-card reader in my A4000T so I figured adding one of these would be beneficial to my Amithlon setup as well. I actually replaced the tray-loading slim drive in my A4000T with a slot-loading one so I decided to do the same here. Slot-loading slim DVD-RW, 3.5″ Multi-card USB reader, both in a 5.25″ bracket, replaced the 5.25″ DVD-RW drive that had been there. I had an open USB header on the motherboard so connected the multi-card reader to it, attached the IDE cable to the slim to IDE adapter that was attached to the drive and booted up. I had to add the additional ports in Poseidon (AmithlonUSB.device 4 and 5) and all was good.
I had installed BB3&4 package earlier and felt everything was good until I tried to launch Shapeshifter….which caused and immediate “emulator crash…” reboot of Amithlon. It’s either something being loaded as a module to replace the kickstart modules or something in the ‘Amiga OS ROM update” file that the BB3&4 skips. I also didn’t see any advantage BB3&4 was adding to my system so I wiped it once more and started the install again from a blank HD. This time I installed with the intent of getting a good base setup that I can restore if needed. So, I installed Amithlon’s OS 3.9 (which includes BB1), AOSXLUpdate1, BB2, AmithlonUpdate131, MUI 3.8, Poseidon/Arakattack, Picasso96 2.1e, Roadshow and icon.library and layers.library + loadmodule. With this all functioning I copied the contents to another partition I created named Backups: I’ll hide this with the Workbench prefs program which only hides the device to Workbench, you still have access from the shell. If something goes crazy from either me experimenting with software patches or libraries, etc…I can easily wipe and copy my “backup” back to the SYS: volume in seconds.
So, new slot loading DVD-RW drive with a Multi-card USB reader, all functioning with a backup made to quickly and easily setup again if I screw things up.
*Shapeshifter now has a real Macintosh feel, you drag the disc icon over the trash and it pops the disc out instead of opening the disc tray. A small bonus with the slot-load drive 🙂
Finally, i686be-amithlon-ahiusr.lha installed and functioning. I just wanted to get it working vs. the m68k-amigaos-ahiusr.lha version. More of a challenge I guess since I kept hearing it was much more difficult to do than the 68k version.
Bernie stated there really isn’t much of a difference, the i686 will be ever so slightly faster, but use it for bragging rights or something Faster because you are running x86 native code vs. m68k emulated code.
You need run_elf in your user-startup to let it interpret the x86 binary.
run <>NIL: run_elf GUARD PATCH
I’ll add the note to my guides and include run_elf 1.8c.tar.gz file.
You’ll need both run_elf and run_elf86 copied to C: or C:amithlonc/
After numerous attempts, finally finding the right components to make it work feels good.
This weekend I formatted my system volume and installed Amithlon again. Leaving all my other partitions alone provided access quickly to most of my installed software, just needed to spend time setting up my Workbench and patching/updating. I also had a copy of my Sys: partition on my DATA: volume in a folder so I could easily pull patch files, libraries, review my startup-sequence, etc… as I was setting up my new install.
This time I skipped doing any Genesis install and I went straight for RoadShow for the TCP/IP stack. Everything is mostly working, I have a few software packages to install but before I continue I may do another fresh install. I want to get Amithlon installed, all the proper updates/patches (AOSXL_update1, Boing Bag2, AmithlonUpdate131) installed, USB and networking configured and Picasso96 installed and configured, then make a full backup of this configuration. Experimental patches and software installs come later as this way I can have a full configured base system (AmigaOS and hardware) to restore if something happens and I can re-evaluate what additional patches and/or software to install from that base.